Development of a novel age-specific pediatric trauma score

D e v e l o p m e n t of a Novel Age-Specific Pediatric Trauma Score By Douglas A. Potoka, Laura C. Sehall, and Henri R. Ford

Pittsburgh, Pennsylvania

Background/Purpose: Trauma scoring systems are needed to provide efficient triage of injured patients and to assess differences in outcomes and quality of care between different trauma centers, Current scoring systems used in pediatric trauma are not age specific, and thus have significant limitations. Methods: The authors queried the Pennsylvania Trauma Outcome Study for all children 0 to 16 years entered in the database from 1993 to 1996. Age-specific threshold values for systolic blood pressure, pulse, and respiratory rate were established. Using coded scores for these age-specific values and Glasgow Coma Scale, an age-specific pediatric trauma score (ASPTS) was derived, Triage ASPTS (T-ASPTS) consisted of the integer sum of coded scores for the 4 variables, whereas ASPTS was calculated using weighted coefficients derived from logistic regression for each variable.

Results: T-ASPTS correlated with mortality rate. Using a

p

pEDIATRIC TRAUMA is the leading cause of death and long-term disability in children. Over 1.5 million childhood injuries occur annually, resulting in approximately 500,000 hospitalizations. ~,2 Moreover, nearly 25% of major injuries occur in patients below 18 years of age, and between 15,000 and 20,000 children die of trauma-related injuries each year. 1,3 There are important physiologic differences between children and adults that may affect the clinical presentation as well as the management of specific injuries. 4,5 The optimal care of injured children requires the proper triage of the most severely injured children to the appropriate regional trauma centers. Thus, trauma scoring systems have been developed to aid in the prehospital evaluation and triage of injured patients. This approach requires an accurate method of assessing injury severity and the physiologic response to injury in the field. In addition to triage, trauma scoring From the Department of Surgery, University of Pittsburgh and Children's Hospital of Pittsburgh, and the Department of Biostatisticw, University of Pittsburgh, Pittsburgh, PA. Presented at the 31st Annual Meeting of the American Pediatric" Surgical Association, Orlando, Florida, May 25-29, 2000. Supported in part by the Children's Hospital Pittsburgh. Address reprint requests to Henri R. Ford, MD, Children's Hospital of Pittsburgh, 3705 Fifth Ave, Pittsburgh, PA 15213. Copyright © 2001 by W.B. Saunders Company 0022-3468/01/3601-0018503. 00/0 doi: l O.l O53/jpsu.2001.20023 106

threshold score of less than 10, T-ASPTS predicted mortality rate with a sensitivity of 96.97% and a specificity of 88.83%. T-ASPTS predicted mortality rate and percentage of patients with Injury Severity Score greater than 20 with similar sensitivity to the Revised Trauma Score (RTS), but T-ASPTS was more specific. The ASPTS predicted probability of survival more accurately than the RTS.

Conclusions: ASPTS performs favorably as both a triage score and as a tool for predicting probability of survival for outcomes analysis. Further comparisons to existing trauma scores are needed to verify the utility of ASPTS. J Pediatr Surg 36:106-112. Copyright © 2001 by W,B. Saunders Company.

INDEX WORDS: Pediatric trauma, trauma scoring systems, revised trauma score, pediatric trauma score.

systems have been developed to predict probability of survival to provide quality assurance at individual trauma centers and to compare outcomes between different trauma centers. A number of trauma scoring systems have been applied to the pediatric trauma population. The Revised Trauma Score (RTS) is a physiologic score based on Glasgow Coma Scale (GCS), systolic blood pressure (SBP), and respiratory rate (RR). 6 One potential limitation of the RTS when applied to children is that the variables are derived from primarily adult data. The Pediatric Trauma Score (PTS) was developed as a triage tool specifically for children, v The PTS combines 6 physiologic and anatomic variables including patient size, airway status, SBP, central nervous system (CNS) status, presence of an open wound, and presence of fracture. 7 PTS has been shown to correlate with injury severity in pediatric patients. 7,8 However, the PTS is rather complex, with 6 variables, some of which may be subjective and poorly defined. When compared directly to the RTS in the triage of injured children, the PTS showed no statistical advantage. 9 Thus, the ideal pediatric trauma scoring system has yet to be defined. A scoring system based on objective age-specific physiologic criteria may be more accurate in predicting injury severity in children than the RTS or PTS. The Pennsylvania Trauma Outcome Study (PTOS) registry was used to derive an age-specific Journal of Pediatric Surgery, Vol 36, No 1 (January), 2001: pp 106-112

AGE-SPECIFIC PEDIATRIC TRAUMA SCORE

107

Table 1. Age-Specific Pediatric Trauma Score GCS

SBP

Pulse

RR

Coded Score

14-15

Normal

Normal

Normal

3

10-13

Mild to moderate hypotension

Tachycardia

Tachypnea

2

4-9

(SBP < mean - 2SD) Severe hypotension

(pulse > mean + SD) Bradycardia

(RR > mean + SD) Hypoventilation

1

3

(SBP < mean - 3SD) 0

(pulse < mean 0

(RR < mean - SD) 0 or intubated

0

SD)

NOTE. Age-specific variables (SBP, pulse, and RR) and GCS were stratified by degree of severity and coded values (0 to 3) were assigned to each variable. The T-ASPTS is the sum total of coded values for all 4 variables.

pediatric trauma score (ASPTS) based on age-defined SBP, pulse, RR, and GCS. The ASPTS was compared with the RTS both as a triage score and as a coded score to predict probability of survival. MATERIALS AND METHODS The PTOS registry contained data on 14,284 children, ages 0 to 16 years, treated at accredited trauma centers in Pennsylvania from 1993 to 1997. A total of 431 burn patients were excluded from the analysis. The population was divided into a study data set consisting of patients entered from 1993 to 1996, and a test data set consisting of patients entered during 1997. The study data set included 9,730 patients with complete data required for analysis, whereas the test data set included 2,248 patients with complete data. Demographic and clinical variables including age, sex, mechanism of injury (blunt versus penetrating), mean Injury Severity Score (ISS), mean GCS, and overall mortality rate were examined for both the study data set and the test data set. The study data set was used to derive the ASPTS. Means and standard deviations were determined for SBP, pulse, and RR for survivors from age 0 to 16 years. The means and standard deviations were then used to divide each physiologic variable (SBP, pulse, and RR) into 4 intervals. GCS intervals were modified from intervals used for the RTS. The intervals defined for each variable and the coded scores (0-3) assigned to each interval are displayed in Table 1. The T-ASPTS was defined as the integer sum of the coded values. ASPTS was evaluated based on its ability to predict both mortality rate and patients with Injury Severity Score (ISS) greater than 20. ISS greater than 20 was chosen as the cutoff to define severe injury. For each integer value of T-ASPTS, mortality rate and percentage of patients with ISS greater than 20 were determined. Using these values, sensitivity and specificity analyses were performed for both mortality rate and ISS > 20. Sensitivity and specificity were determined using T-ASPTS levels of less than 8, less than 9, less than 10, or less than 11 to determine triage to a regional trauma center. Based on this analysis, a T-ASPTS of less than 10 was chosen as the criteria to define severe injury, and, thus, a need for triage to a regional trauma center. This threshold value was used for the subsequent analysis and comparison with the RTS using the test data set. The ASPTS was verified and compared with the RTS using the test data set. Using the test data set, mortality rate and percentage of ISS greater than 20 was determined for each T-ASPTS and RTS score. An RTS less than 12 has previously been proposed as the RTS level to define severe injury. 6 Sensitivity and specificity were then determined for both the T-ASPTS and the RTS using T-ASPTS less than 10 and RTS less than 12 to define severe injury and mortality risk. To derive an age-specific score to be used for outcome analysis,

ASPTS was calculated as the weighted sum of the coded values for each variable (GCS, SBP, pulse, and RR) according to the formula: ASPTS = W~GCSc + WzSBP~ + W3PULSE~ + W4RR~ Where W represents weights derived from logistic regression using the study data set. The subscript "c" indicates that coded values based on Table 1 were used. For each ASPTS value, probability of survival was determined using the logistic function: Ps = 1/(1 + e A) Where P~ is the probability of survival, e is the base of the natural logarithm system, and A = W0 + W1GCS~ + W2SBPc + W3PULSEc + W4RRc. The Ws are the regression weights corresponding to each variable (GCS, SBP, PULSE, and RR) and W o is included to provide for estimation of an intercept in the logistic model. 1° Using the test data set, probability of survival as calculated by ASPTS was compared with actual survival rate. To compare probability of survival calculated from ASPTS to that predicted by the RTS, mean P~ for five P~ intervals (.00 to .10, .11 to .40, .41 to .60, .61 to .90, and .91 to 1.00) derived from the study data set were used to calculate the expected number of survivors at each Ps interval for both the ASPTS and the RTS. At each P~ interval, the expected number of survivors was compared with the actual number of survivors to derive an observed to expected ratio. At each Ps interval, the 95% confidence interval was calculated for the observed to expected ratio to determine significance. SAS statistical software (SAS Institute, Cary, NC) was used to obtain summary statistics, including means and standard deviations for all continuous variables.

RESULTS Table 2 shows demographic variables for both the study data set and the test data set. The study and test data sets were similar with regard to patient age, sex, Table 2. Comparison of Demographic and Clinical Variables Between the Study and Test Data Sets Variable

Study Data Set (1993 to 1996)

Test Data Set (1997)

Number Mean age (yr) Sex (% male) Mechanism (% blunt) Mean ISS Mean GCS Mortality rate (%)

9,730 9.3 ± 5.2 68.4 88.8 11.2 ± 10.6 13.3 ÷ 3.7 4.4

2,248 9.5 ± 5.1 65.5 90.3 11.9 ± 11.1 13.3 ± 3.8 4.0

108

POTOKA, SCHALL, AND FORD

Systolic Blood Pressure ~ - -

140

~120 1

EE~10o~ _ ~

..

--

--_ -- -- --

8O

~o 8°1 ! 40

= O

~ "l"rnean - - I " mean - 2SD

2""-'~ao- ~ L! l 0

A

o

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

Age (years)

Pulse

2oo~ 180

t~ m e a n

I~

-

140

-

~

~120

=,00 8o

60

- - - -

40

1

!1 I

; - I - . . . . . . SDi

i

- .

20

.

.

.

J

o

L

B

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

Age (years)

Respiratory Rate 45-~

~ - - O " - rnean I --B--mean + Sb ~'-i"-rnean- SD ~ ,

-

40

O 25

~

~

~o~ ,

,,

._

._

I

thresholds for the coding of 4 variables: GCS, SBP, pulse, and RR. Systolic blood pressure, pulse, and RR scoring thresholds were determined for each age based on the survivor means and standard deviations. For each variable, coded scores were tabulated according to the data in Table 1. Actual threshold values for SBP, pulse, and RR at each age are displayed in Fig 1. The T-ASPTS is defined as the integer sum of the individual coded scores for each of the 4 variables: GCS, SBP, pulse, and RR. The range of the T-ASPTS includes integer values from 0 to 12. For example, based on the data in Fig 1 and Table 1, a hypothetically injured 2 year old with a SBP of 80 (3), a pulse of 120 (3), a respiratory rate of 32 (3), and a GCS of 8 (1) would have a T-ASPTS of 10. The T-ASPTS correlated with mortality rate and the percentage of children with ISS greater than 20 in the study data set (Table 3). There was a roughly linear increase in mortality rate as T-ASPTS dropped from 9 to 3. Patients with scores below 3 had 100% mortality rate, whereas those with scores above 9 had almost 100% survival rate (Table 3 and Fig 2). The T-ASPTS was associated with a similar trend of percentage of children with ISS greater than 20, although this appeared to be less strongly correlated than mortality rate. To determine which threshold value of T-ASPTS should be used to determine the risk of increased mortality and need for transfer to a regional trauma center, sensitivity and specificity analyses were performed. Sensitivity and specificity for predicting both mortality and ISS greater than 20 were determined for T-ASPTS for threshold values of less than 8, less than 9, less than 10, and less than 11 (Table 4). Based on this analysis, a T-ASPTS threshold value of less than 10 appeared optimal for predicting mortality, with sensitivity of 98.84% and specificity of 89.02%. In contrast, a T-ASPTS threshold value of less than 10 only had a sensitivity of 68.06% and a specificity of 75.18% for predicting ISS greater than 20. For each threshold value, the T-ASPTS Table 3. Correlation of T-ASPTS With Mortality Rate and ISS Greater Than 20 Integer

f

C

t

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16

Age (years) Fig 1. Age-specific means and threshold values derived from the study data set used to define ASPTS intervals for systolic blood pressure (A), pulse (B), and respiratory rate (C).

mechanism of injury, mean ISS, mean GCS, and overall mortality rate. The male to female ratio was roughly 2:1; and blunt trauma accounted for approximately 90% of injuries. The study data set was used to define age-specific

Score

No. of Patients

Mortality Rate (%)

ISS > 20 (%)

0 1 2 3 4 5 6 7 8 9 10 11 12

152 16 7 31 91 198 325 167 183 276 1539 1381 5364

100.00 100,00 100.00 83.37 54.95 28.28

63.82 62.50 57.14 67.74 72.53 58.59 55.38 54.49 40.98 36.59 8.84 10.93 5.03

20.62

11.38 9.29 5.43 0.32 0.00 0.00

AGE-SPECIFIC PEDIATRIC T R A U M A SCORE

T - A S P T S vs

109

Mortality

T a b l e 5. C o m p a r i s o n

of Sensitivity and Specificity for T-ASPTS

and

RTS Using Test Data Set

Trauma Score

Sensitivity and Specificity for Mortality

Sensitivity and Specificity for ISS > 20

T-ASPTS ~ 10

Sensitivity = 96.97%

Sensitivity - 49.43%

6o

RTS < 12

Specificity - 88.83%

Specificity = 91.83%

Sensitivity = 100.00%

Sensitivity - 64.86%

Specificity

Specificity = 77.71%

74.05%

3o 20

88.83% and 74.05%, respectively. When predicting ISS greater than 20, sensitivity for both T-ASPTS (49.43%) and RTS (64.86%) were reduced. Again, the T-ASPTS was more specific in predicting ISS greater than 20 compared with the RTS, with specificities of 91.83% and 77.71%, respectively. To derive a score to be used for outcome analysis, ASPTS was calculated as the weighted sum of the coded values for each variable (GCS, SBP, pulse, and RR) according to the formula:

1 A

0

2

~

4

s 0 • T-ASPTS

8

9

10

i1

T - A S P T S vs I S S > 2 0

ASPTS = W1GCS~ + W2SBP~ + W3PULSEc + W4RRc

B

o

1

2

3

4

5 s 7 T-ASPTS

s

9

10

Fig 2. C o r r e l a t i o n of T - A S P T S w i t h ( A ) m o r t a l i t y a g e of p a t i e n t s w i t h ISS g r e a t e r t h a n 20.

11

1~

a n d (B) p e r c e n t -

was more sensitive in predicting mortality than ISS greater than 20, whereas specificities were similar for mortality and ISS greater than 20. To verify the T-ASPTS and to compare it with the RTS, the test data set was utilized to calculate sensitivities and specificities for predicting both mortality and ISS greater than 20 for each score (Table 5). A threshold of less than 10 was chosen for the T-ASPTS as determined by the previous analysis using the study data set. A threshold score of less than 12 has been proposed previously for using the RTS as a triage tool. 6 For predicting mortality, the T-ASPTS and RTS had similar sensitivities of 96.97% and 100%, respectively. However, the T-ASPTS was more specific in predicting mortality compared with the RTS, with specificities of T a b l e 4. S e n s i t i v i t y Mortality

and Specificity of T-ASPTS

as a P r e d i c t o r o f

a n d ISS G r e a t e r T h a n 2 0 b y T h r e s h o l d

Level

The Ws represent weights derived from logistic regression using the study data set: W 1, 1.8945; W 2, 1.4366; W 3, 0.5908; W 4, 0.1843. The subscript "c" indicates that coded values based on Table 1 were used. The ASPTS is a continuous score ranging from 0 to 12.32. ASPTS was analyzed using the test data set. For each ASPTS interval, the mean probability of survival was determined and compared to the actual survival rate (Table 6). There was a strong correlation between predicted probability of survival and actual survival rate in each ASPTS interval except for the interval 1.00 to 1.99 where there was only 1 patient who had a probability of survival of 0.0264 but lived. The test data set was used to compare probability of survival calculated from ASPTS with that predicted by T a b l e 6. P r o b a b i l i t y of S u r v i v a l B a s e d o n A S P T S

ASPTS Interval

No. of Patients

Probability of Survival (Ps)

Actual Survival Rate

0.00-0.99

33

0.0089

0.00

1.00-1.99

1

0.0264

100.00

2.00-2.99

5

0.0904

0.00

3.00-3.99

4

0.1873

25.00

4.00-4.99

23

0.4953

47.83

5.00-5.99

49

0.6669

79.59

6.00-6.99

96

0.8093

83.33

7.00-7.99

65

0.9459

92.31

8.00-8.99

38

0.9760

97.37

52

0.9917

94.23

T-ASPTS Threshold

Sensitivity and Specificity for Mortality

Sensitivity and Specificity for ISS > 20

T-ASPTS < 8

Sensitivity - 91.40%

Sensitivity = 44.39%

Specificity = 93.61%

Specificity = 95.22%

10.00-10.99

103

0.9966

99.03

T-ASPTS < 9

Sensitivity - 95.35%

Sensitivity = 50.08%

11.00-11.99

462

0.9986

100.00

Specificity - 91.83%

Specificity = 93.64%

12.00-12.32

1317

0.9995

99.92

T-ASPTS < 10

Sensitivity - 98.84%

Sensitivity = 68.06%

Specificity = 89.02%

Specificity = 75.18%

T-ASPTS < 11

Sensitivity = 100.00%

Sensitivity = 79,51%

Specificity - 72.53%

Specificity - 60.56%

9,00-9.99

NOTE. ASPTS intervals are based on scores calculated using logistic regression as described. Estimated probability o f survival was calculated as described in the M e t h o d s section and c o m p a r e d w i t h actual survival.

POTOKA, SCHALL, AND FORD

110

Table 7. Comparison of Probability of Survival Determined by ASPTS Versus RTS Ps Interval ASPTS

RTS

Mean Ps

No. of Patients 36 9 12

OBS

EXP

1 1 13

0 2 12

OBS/EXP

95% CI

-.50 1.08

.04-16.36 0.01-2.78 .58-1.85

.00-. 10 •11 -.40 .41 -.60

0.01 0.22 0.52

.61 -.90

0.77

110

117

110

1.06

.88-1.27

.91-1.00 .00-.10 •11-.40

0.99 0•03 0.31

2017 35 137

2026 1 98

2017 1 42

1.00 1.00 2.33

.95-1.05 .01-5.56 1•89-2.84*

.41 -.60 .61 -.90

0.54 0.76

29 78

21 73

16 59

1.31 1.24

.81-2.01 .97-1.56

.91-1.00

0.98

1969

1965

1930

1.02

.97-1.05

NOTE• Mean probability of survival (Ps) was determined for P~ intervals using the study data set. Mean Ps values were then applied to the test data set to calculate expected survival at each Ps interval. Expected survival (EXP) was compared with observed survival (OBS) to derive observed to expected (OBS/EXP) ratios and 95% confidence intervals (95% CI). *Statistically significant difference between observed and expected at the c~ = 0.05 level.

the RTS (Table 7). Observed to expected ratios were calculated based on the ASPTS and the RTS for 5 Ps intervals (.00 to .10, .11 to .40, .41 to .60, .61 to .90, and .91 to 1.00) derived from the study data set. For each Ps interval, the 95% confidence interval was calculated for the observed to expected ratio to determine significance. Probability of survival calculated using the ASPTS accurately predicted survival at each Ps interval. The RTS did not accurately predict survival in the . 11 to .40 Ps interval (observed to expected 95% CI 1.89 to 2.84). Furthermore, for P~ intervals .41 to .60, and .61 to .90, the RTS observed to expected ratios were higher with wider 95% confidence levels, but these did not reach statistical significance. DISCUSSION

Pediatric trauma is a significant health problem in the United States with over 1.5 million injuries, 500,000 hospitalizations, and approximately 15,000 to 20,000 deaths in children yearly caused by trauma. 1,2 Pediatric trauma centers and adult trauma centers with a strong commitment to pediatric trauma have been established to serve as regional resources in the management of severely injured children. The large number of injured children compared with the relative scarcity of these regional resource centers for pediatric trauma dictates that not all injured children can be treated in these settings. Therefore, the proper utilization of these limited resources depends on the triage of the most severely injured children to these centers, while avoiding the overtriage of children with minor injuries. Successful triage requires an accurate pediatric trauma scoring system to identify severely injured children. Several trauma scoring systems have been used in the pediatric population. 11 The RTS is a physiologic score based on GCS, SBP, and RR. 6 The RTS values for SBP and RR are based on adult values, which may

potentially limit the effectiveness of the RTS in children; especially younger children who are physiologically different than adults. Despite these limitations, the RTS has been verified in children in several studies. 12J3 The PTS was developed as a triage tool specifically for children. 7 The PTS combines 6 physiologic and anatomic variables including patient size, airway status, SBP, CNS status, and presence of an open wound or a fracture(s). 7 The PTS has been shown to correlate with injury severity in pediatric patients. 7,8 However, with 6 variables, the PTS is rather complex, and some of the variables may be subjective and poorly defined. For example, the "open wound" is poorly defined, whereas the CNS and airway variables may be subjective. When compared directly with the RTS in the triage of 376 injured children, ages 0 to 14, the PTS showed no statistical advantage over the RTS.9 We have developed a new trauma score for injured children. The ASPTS has 2 potential advantages over existing scores. First, the threshold values for the 3 physiologic variables (SBP, pulse, and RR) are age specific and based on the mean values for survivors in our study set population. The use of age-specific physiologic variables should theoretically improve the accuracy of the ASPTS compared with trauma scores derived in adults and applied to children. Second, it is based on objective physiologic criteria and, therefore, may be easier to score and can be calculated from data available in the field or retrospectively from charts or computerized databases. Thus, the ASPTS potentially can be useful for both triage and outcome analysis. The ASPTS was derived using a study data set of 9,730 children, ages 0 to 16 years, treated at accredited trauma centers in the state of Pennsylvania, and entered in the PTOS registry from 1993 to 1996. This subset had a mean age of 9 years and a roughly 2:1 male to female ratio. Blunt trauma accounted for

AGE-SPECIFIC PEDIATRIC TRAUMA SCORE

88.8% of injuries in this population. Minor and moderate injuries predominated with a mean ISS of 11 and a mean GCS of 13. These findings are consistent with other reported series of pediatric trauma. 14-17 Thus, the ASPTS should be applicable to most pediatric trauma populations. In a manner similar to the RTS, a triage ASPTS was derived as the integer sum of the score (0 to 3) for each of the 4 variables. This integer score ranges from 0 to 12. As seen in Table 3, the T-ASPTS correlated strongly with mortality rate in the study population. Mortality rate was 100% for a score less than 3, and 0% for a score greater than 10, with a roughly linear distribution between 3 and 10. We also chose ISS greater than 20 as an indicator of severe injury. The T-ASPTS also correlated with ISS greater than 20, although less strongly than with mortality (Table 3). Sensitivity and specificity analyses were performed at ASPTS cutoff points of 8, 9, 10, and 11 to determine the optimal threshold point for further analysis. Based on this analysis, an ASPTS score below 10 was chosen as an indicator of severe injury requiring triage to a regional pediatric trauma center. This score predicted mortality rate with a sensitivity of 98.84% and a specificity of 89.02%. It was less optimal for predicting ISS greater than 20, with a sensitivity of 68.06% and a specificity of 75.18% when applied to the study data set. The T-ASPTS using a threshold value of less than 10 as an indicator of severe injury was applied to the test data set of 2,248 patients and compared directly with the RTS. Previously, an RTS of less than 12 had been chosen as an indicator of severe injury. 6 Both T-ASPTS and RTS predicted mortality with high sensitivity, 96.97% and 100%, respectively. However, the specificity of the ASPTS for predicting mortality (88.83%) was superior to that of the RTS (74.05%). For predicting ISS greater than 20, the sensitivities were lower for both the T-ASPTS and the RTS, 49.43% and 64.86%, respectively. However, the specificity of the ASPTS was 91.83% compared with 77.71% for the RTS. The T-ASPTS had a sensitivity of only 49.43% for predicting ISS greater than 20 in the test data set suggesting that the T-ASPTS misses more severely injured patients with ISS greater than 20. However, the sensitivity of 96.97% for predicting mortality suggests that the most severely injured children (with high mortality rates) are indeed identified by the T-ASPTS. The discrepancy between sensitivity for predicting mortality compared with ISS greater than 20 suggests that ISS alone may not be adequately sensitive to detect severely injured children without a

111

concurrent physiologic assessment. In addition to predicting mortality with high sensitivity, the T-ASPTS predicts both mortality and ISS greater than 20 in children with higher specificity compared with the RTS. These properties should reduce overtriage of children with minor injuries to regional pediatric trauma centers and thus provide more efficient utilization of resources. Unfortunately, we were unable to compare the ASPTS to the PTS using the National Pediatric Trauma Registry (NPTR) because of missing variables in the registry. A potential drawback of the ASPTS for triage is that it utilizes different values based on age for scoring each physiologic variable. Determining the ASPTS in the field would require either a knowledge of the normal and abnormal vital signs throughout infancy and childhood, or reliance on tables or graphs to determine normal and abnormal ranges for each variable. However, if this knowledge is available, determining the T-ASPTS simply requires assessment of the child's mental status (GCS); respiratory status (ie, whether the child is tachypneic, hypoventilating, or intubated); the heart rate (tachycardia or bradycardia); and blood pressure (mild or severe hypotension). For use in outcome analysis, a coded ASPTS also was derived using logistic regression. This method resulted in a continuous score ranging from 0 to 12.32. The coded ASPTS can be used to calculate probability of survival. Probability of survival determined by ASPTS accurately predicted actual survival in the test data set. The RTS also can be used to calculate probability of survival in a similar manner. In fact, RTS is a component of TRISS, which has been used to predict survival in both adult and pediatric trauma populations, a* Therefore, we also calculated probability of survival using RTS in the test data set. There was a trend for RTS to underestimate actual survival in the 0.11 to 0.90 probability of survival range, although there was a statistically significant difference only for the 0.11 to 0.40 probability of survival interval (Table 7). Therefore, other than at the highest and lowest probabilities of survival where both ASPTS and RTS perform similarly, ASPTS appears to predict survival more accurately in injured children. This observation suggests that ASPTS may be a better tool for predicting survival in injured children and for outcomes analysis. It also suggests that TRISS methodology would be more accurate using ASPTS rather than RTS when applied to injured children. We have shown that the T-ASPTS is highly sensitive in predicting mortality in pediatric trauma patients and is more specific than RTS in identifying severely injured children. Furthermore, the ASPTS also predicts survival

112

POTOKA, SCHALL, AND FORD

in injured children accurately. H o w e v e r , a n e w scoring system should p r o v i d e substantial advantage o v e r the existing scoring systems to justify its use. Further c o m parisons with existing scoring systems, including the

PTS, w h i c h we w e r e unable to calculate f r o m the P T O S database, and with larger n u m b e r o f patients are n e e d e d to d e t e r m i n e if the A S P T S p r o v i d e s a distinct advantage o v e r existing scoring systems.

REFERENCES 1. Schafermeyer R: Pediatric Trauma. Emerg Med Clin North Am 11:187-205, 1993 2. Sanchez JI, Paidas CN: Childhood trauma: Now and in the new millenium. Surg Clin North Am 79:1503-1535, 1999 3. Rivera FP: Pediatrc injury control in 1999: Where do we go from here? Pediatrics 103:883-888, 1999 4. Tepas JJ, DiScala C, Ramenofsky ML, et al: Mortality and head injury: The pediatric perspective. J Pediatr Surg 25:92-96, 1990 5. Powell M, Courcoulas A, Gardner M, et al: Management of blunt splenic trauma: Significant differences between adults and children. Surgery 122:654-660, 1997 6. Champion HR, Sacco WJ, Copes WS, et al: A revision of the trauma score. J Trauma 29:623-629, 1989 7. Tepas JJ, Mollitt DL, Talbert JL, et al: The pediatric trauma score as a predictor of injury severity in the injured child. J Pediatr Surg 22:14-18, 1987 8. Tepas JJ, Ramenofsky ML, Mollitt DL, et al: The pediatric trauma score as a predictor of injury severity: An objective assessment. J Trauma 28:425-429, 1988 9. Kaufmann CR, Maier RV, Rivara FP, et al: Evaluation of the pediatric trauma score. JAMA 263:69-72, 1990 10. Walker SH, Duncan DB: Estimation of the probability of an

event as a function of several independent variables. Biometrika 54: 167-79, 1967 11. Furnival RA, Scbunk JE: ABCs of scoring systems for pediatric trauma. Pediatr Emerg Care 15:215-223, 1999 12. Eichelberger MR, Gotschall CS, Sacco WJ, et al: A comparison of the Trauma Score, the Revised Trauma Score, and the Pediatric Trauma Score. Ann Emerg Med 18:1053-1058, 1989 13. Eichelberger MR, Bowman LM, Sacco WJ: Trauma Score versus Revised Trauma Score in TRISS to predict outcome in children with blunt trauma. Ann Emerg Med 18:939-942, 1989 14. D'Amelio LF, Hammond JS, Thomasseau J, et al: "Adult" trauma surgeons with pediatric commitment: A logical solution to the pediatric trauma manpower shortage. Am Surg 61:968-974, 1995 15. Fortune JB, Sanchez J, Grace L, et al: A pediatric trauma center without a pediatric surgeon: A four-year outcome analysis. J Trauma 33:130-139, 1992 16. Rhodes M, Boorse SSD: Pediatric trauma patients in an 'adult' trauma center. J Trauma 35:384-393, 1993 17. Nakayama DK, Copes WS, Sacco W: Differences in trauma care among pediatric and nonpediatric trauma centers. J Pediatr Surg 27: 427-431, 1992 18. Boyd CR, Tolson MA, Copes WS: Evaluating trauma care: The TRISS method. J Trauma 27:370-378, 1987

Discussion J. Tepas (Jacksonville, FL): I h a v e a n u m b e r o f c o m ments and 50 questions, but I will shorten it down. W e d e v e l o p e d the pediatric trauma score w h e n A T L S was first in evolution, and its purpose was to reiterate the concept o f A T L S to l o o k at specific variables. W e then were c o n v i n c e d to put numbers on it and w e r e a m a z e d to find out that the prehospital p e o p l e did not like to add. I do not k n o w if that is a c o m m e n t about our A m e r i c a n education or not. H e r e today in Florida w e actually do h a v e by Florida statutes a c o l o r - c o d e d PTS, which, if they can r e c o g n i z e colors, they can use it to triage. M y questions are 3, h o w e v e r . The first is, death is a very small variable unfortunately, and it is a finite one. H a v e you l o o k e d at this in relation to any o f the other real o u t c o m e measures, specifically length o f stay, v e n t days, G O S and things o f that variety? The second question is, in d e t e r m i n i n g y o u r c o d e d score, you used essentially a logistic regression with a W a l k e r D u n c a n analysis and those codes, those variables h a v e n e v e r been s h o w n to be transportable f r o m one

patient population to the next, and there is a w h o l e raft o f literature that d o c u m e n t s that w h a t w o r k s in W a s h i n g t o n does not w o r k in N e w Y o r k does not w o r k in England. So, what are y o u r plans to test this both in the field as fax" as acceptability by the prehospital providers or potentially referring hospitals? Second, where are you going to take it out of the s o m e w h a t u n i q u e p r o g r a m in Pennsylvania to other less well o r g a n i z e d systems? D.A. Potoka (response): T h e only other variable w e e x a m i n e d with our score was the ISS. W e s h o w e d similar predictability for ISS greater than 20 with our score c o m p a r e d with the R T S . W e h a v e not e x a m i n e d other o u t c o m e variables yet, although w e plan to do that. As far as the applicability to other populations, w e do plan on validating this versus both the R T S and the PTS in another data base. W e h a v e acquired N P T R data, although we h a v e been unable to do that analysis at this point because there w e r e quite a f e w missing variables for b l o o d pressure, pulse, and heart rate in the subset o f the N P T R that w e acquired so far, but w e do plan on doing that in the future.